View
1
Download
0
Category
Preview:
Citation preview
1
Scouting for good jobs:
Gender and networking in job search
Elena Obukhova
McGill University
Desautels Faculty of Management
elena.obukhova@mcgill.ca
514.893.3896
Adam M. Kleinbaum
Dartmouth College
Tuck School of Business
adam.m.kleinbaum@tuck.dartmouth.edu
603.646.6447
September 26, 2018
Keywords: Networking; social networks; gender; labor markets.
This paper benefitted from discussions with and comments from Emily Bianchi, Dan Cable, Tiziana Casciaro, Lisa
Cohen, Laura Doerring, Roberto Fernandez, Isabel Fernandez-Mateo, Connie Helfat, Matissa Hollister, Martin Kilduff,
Jen Merluzzi, Siobhan O’Mahony, Brian Rubineau, Sandra Smith, Wendy Smith, Olav Sorenson, Adina Sterling, and
Peter Younkin; seminar participants at HEC Paris, London Business School, the University of Chicago, Yale
University and members of the Montreal Social Networks Working Group; students in the Tuck Social Networks in
Organizations seminar; and conference participants at the Stanford Hiring and Organizations Conference, the
European Group for Organizational Studies, the Junior Faculty Workshop in Organization Theory, the Academy of
Management annual meetings, and the Sunbelt Social Networks conference. We also gratefully acknowledge the many
and various staff members at the university studied for their support of the data collection. The usual disclaimer applies.
2
Scouting for good jobs:
Gender and networking in job search
While networking – or the purposeful creation of new social ties for professional goals – is widely
seen as a key to success in job search, prior research does not make clear predictions about whether
women’s and men’s networking strategies might differ. In this paper we clarify these theoretical
predictions and propose a new mechanism that we term “scouting,” by which women network –
especially with other women – to “scout” potential employers, or to gain an in-depth understanding
of firms’ organizational culture and practices, especially as they might affect a prospective female
employee. To provide data for this new mechanism, we leverage a unique research setting and
examine networking outreach among similarly qualified men and women presented with a similar
pool of potential contacts. Specifically, we study networking with alums by job-seeking students
in an elite MBA program using server logs to directly observe students’ outreach behavior.
Consistent with “scouting,” we find that female students reach out to significantly more women
and to at least as many men as their male classmates. We discuss the implications of our findings
for our understanding of gender differences in networks and career attainment, especially as they
relate to the “whisper networks” revealed by the #MeToo movement.
3
It is widely understood that contacts play an important role in finding a job (Granovetter
1974; Fernandez, Castilla and Moore 2000; Castilla, Lan and Rissing 2013). Research reveals
that as many as half of jobs in the U.S. are found with a help of social networks (Marsden and
Gorman 2001; Rubineau and Fernandez 2015). Through interactions with contacts, job-seekers
can explore potential career options, determining what type of employer might offer them the
best opportunities to succeed, how to present themselves during an interview or even find
sponsors in the organization (Barbulescu 2015; Rivera 2015; Greenberg and Fernandez 2016).
More recently, the rise of the #MeToo movement, brought to light the existence of “whisper
networks,”1 in which women have been reported to deliberately and strategically share
information about individuals and employers, so as to avoid hostile working environments and
bosses who engage in sexual harassment, making more salient and more important than ever to
understand how job-seekers use contacts to find working environment where they can survive
and thrive.
When people lack the contacts they need for successful job search, they work to acquire
those contacts through networking, the purposeful creation of new social ties for achievement of
professional goals (Sharone 2013; Casciaro, Gino and Kouchaki 2014). Yet the research on
gender differences in job search networking remains nascent, with few clear theoretical
predictions about how women and men might differ in their networking strategies. On the one
hand, a growing literature has focused on understanding gender differences in benefits women
and men derive from networking showing that women receive less benefits from contacts than
men (Huffman and Torres 2002; Abraham 2017) or that women are negatively evaluated for
1 See for example: https://www.newsweek.com/what-whisper-network-sexual-misconduct-allegations-719009;
https://www.nytimes.com/2017/11/04/business/sexual-harassment-whisper-network.html. Accessed September 17th,
2018.
4
networking (Brands and Kilduff 2014). While these studies do not explore how women adjust
their networking strategies in response to receiving less benefits from networking than men do,
one possible implication from this research is that women might “lean out” from networking (c.f.
Brands and Fernandez-Mateo 2018) – that is, network less than men (e.g. Forret and Dougherty
2001).
On the other hand, the existing research women’s functionally differentiated networks
and responses to labor market discrimination and suggests an opposite prediction. Research finds
that women maintain “functionally differentiated networks,” in that they receive distinct
resources from women than from men (Ibarra 1992, 1997). We also know that women adjust
their job search strategy in response to anticipated gender-based barriers in organizations
(Barbulescu and Bidwell 2013; Pager and Pedulla 2015; Sterling 2017). Taken together, these
two bodies of work imply what we call “scouting for good jobs”, a mechanism by which women
use gender-based networks to seek information about the job, the bosses, or the firm more
generally, that can help women to chose an employer where they can survive and thrive
professionally (Shih 2006). The “whisper networks” brought to light the #MeToo movement are
one expression of this more general phenomenon of “differentiated networking” in which women
network with men to receive general networking benefits and network with additional women to
engage in scouting. This latter argument would mean that women engage in more network
outreach to women, which might lead them to network more than men.
Our study seeks to clarify our theoretical expectations about gender differences in
networking strategies and to provide novel empirical evidence to support our theoretical claims.
To examine networking strategies in a field setting and provide evidence consistent with
“scouting for good jobs”, we leverage a strategic research setting that allows us to unobtrusively
5
observe network outreach among similarly qualified men and women who are presented with a
similar opportunity set of potential contacts. Specifically, we study networking with alumni2 by
job-seeking students in an elite MBA program. We use server logs of the school’s alumni
database to directly observe students’ outreach behavior. Importantly, we collect information on
when students reached out to alums, before alums have an opportunity to respond; that is, we do
not restrict our sample to successful networking attempts, which would potentially bias results.
Also, women and men in our setting are presented with an identical pool of potential contacts in
the alumni database, thus minimizing differences in access to potential contacts that might lead
women to network with other women in other settings.
Consistent with a “scouting” mechanism, our results indicate that female MBAs reach out
to more female alums than male MBAs. Importantly, we find that women’s outreach to more
women does not come at the expense of reaching out to resource-rich contacts: we find that
women are no less active than men are in interacting with senior-level alums of either gender or
with male alums. As a result, and contrary to predictions that women might “lean out” of
networking, we find that women network more – and not less – than their male counterparts do.
We test and find little evidence for an alternative explanation of our results that women network
with women because they receive more help from them or because they receive less help from
men. We also offer some evidence from interviews with female MBAs to highlight how they use
networking with other women to engage in “differentiated networking” (cf. Ibarra 1992) in
which they seek information and access resources in the same way that men do, and in addition,
they reach out to women to “scout for good jobs.” The need that women perceive to reach out to
female alumni in addition to all the networking that their male peers do is potentially costly; we
2 For parsimony and gender equity, we use the singular abbreviation “alum” and the plural “alumni” or “alums”
throughout, rather than the more formal, but gender-specific, words “alumnus,” “alumna,” and “alumnæ”.
6
suggest that this cost is a kind of insurance against encountering gender-based barriers in the
workplace.
By leveraging a unique research setting to observe some processes previously invisible to
scholars, this paper contributes to the literatures on social networks and gender in labor markets
and will, we hope, reinvigorate research at their intersection in at least two ways. First, our
results suggest that the “whisper networks” that recent media accounts describe as women’s
response to workplace sexual misconduct, harassment and assault may not be limited to the
Harvey Weinsteins of the world; rather, they are an aspect of the pattern of differentiated
networking and reflect a broader process of women using networks to “scout” for good jobs.
Second, our study also brings to light the gendered process through which people create and use
their social networks, contributing to our understanding of deliberate tie formation through
networking (Casciaro, Gino and Kouchaki 2014). More broadly, our study extends our theories
of the relationship between network access and mobilization (Kwon and Adler 2014), and
structure and agency in network research, more generally (Kilduff, Tsai and Hanke 2006; Ahuja,
Soda and Zaheer 2012; Burt 2012).
RESEARCH ON WOMEN AND NETWORKING
Prior research does not make clear predictions about how women’s and men’s
networking strategies might differ during the job search. One important body of research that
speaks to this question focuses on how others react to women’s networking and might imply that
women “lean out” of networking, because they do not benefit from networking as much as men
do. In this stream of work, one line of research highlights that contacts might not provide as
much help to women as they do to men. For example, in a study of participants in job clubs,
7
where unemployed job seekers exchange information about job leads, Huffman and Torres
(2002) find that women receive worse job leads than men.3 Similarly, in a study of business
referral clubs, Abraham (2017) finds that while women are equally likely to receive business
leads from their contacts, contacts are much less likely to refer women to third-parties, in part
because referrers expect these third-parties to exhibit a gender bias. Although McDonald (2011)
finds that, controlling for network characteristics, men and women receive the same number of
unsolicited job leads from their contacts, he also finds that white men are less willing to vouch
for female than male job candidates.
Another line of research suggests that networking might be less beneficial for women
than for men because due stereotype incongruence women who network are subject to social
sanctions. According to theories of social categorization, networking behaviors are perceived as
agentic and instrumental (Ingram and Zou 2008; Casciaro, Gino and Kouchaki 2014; Kuwabara,
Hildebrand and Zou 2016), and are therefore incongruent with prevailing gender stereotypes,
which suggest that women are cooperative, communal, and nurturing (West and Zimmerman
1987; Heilman et al. 1989; Eagley and Karau 2002); consequently, women who engage in
networking are judged harshly. For example, in a widely-cited, albeit unpublished, classroom
experiment, Flynn (2007) taught a case about a successful venture capitalist, identified to half the
students as Heidi Roizen and to the other half as Howard Roizen. While both groups of students
agreed that Roizen was extremely competent and effective, students who believed they had read
about a man rated him as more genuine, humble, and kind and were more likely to say they
3 Huffman and Torres (2002) also find that women give other women lower wage job leads. Consistent with
scouting, they speculate that “the negative effect for women could be due to the fact that leads offered by women to
other women are for jobs that offer better opportunities for balancing work-family demands… It may be the case
that women have different kinds of conversation with their female contacts that involve more nuanced evaluation of
jobs” (p. 809).
8
would favor hiring Howard. In contrast, students who believed they had read about a woman
were more likely to view Heidi as power-hungry, self-promoting, and disingenuous. Consistent
with this, Brands and Kilduff (2014) find that women who occupy brokerage positions are seen
as incongruent with their stereotyped role and are therefore at risk of social sanction.
One possible implication from this research is that women might “lean out” from
networking, or engage less in networking than men do, instead re-allocating their energies
elsewhere (e.g. Forret and Dougherty 2001). In recounting the Roizen experiment in her book
Lean In, Sheryl Sandberg concludes, “this bias is at the very core of why women are held back.
It is also at the very core of why women hold themselves back,” (Sandberg 2013: p. 40). And
while we know of no direct evidence on whether or not women “hold themselves back” from
networking, evidence from women’s behavior in other areas might suggest that they do. For
example, Merluzzi (2017) argues that such stereotype incongruence accounts for women having
fewer negative ties. In labor market context, Brands and Fernandez-Mateo (2017) find that when
women are a negatively stereotyped minority, after being rejected for a job, women are less
likely than men to re-apply for a job with the same employer. If similar processes apply to
networking, women might also be more sensitive to bias or outright rejection in networking
interactions, leading them to avoid further networking attempts.
However, we also have good reason to question whether women’s different networking
outcomes drive out women from networking. Research on network dynamics suggests that
benefits individual receive through networks do not adequately explain how people create and
utilize their ties (e.g. Ryall and Sorenson 2007; Buskens and van de Rijt 2008). A related stream
of research on network access and mobilization suggests that benefits available through a
person’s network do not explain which ties individuals mobilize (Kwon and Adler 2014: 414).
9
And in the context of job-search Obukhova and Lan (2013) find that job candidates with more
social capital are no more likely to use it to look for jobs, even though those who use it, benefit
from doing so. In sum, these two streams of research suggests that it is important to study
network outreach directly rather than infer it from the pattern of benefits it confers.
In the next section, we draw on research on women’s “functionally differentiated
networks” (Ibarra 1992, 1997) and on women’s responses to labor market discrimination
(Barbulescu and Bidwell 2013; Pager and Pedulla 2015; Sterling 2017), to develop an alternative
prediction: that in addition to all the resources of information and access that men network for in
their job searches, women also network, primarily with other women, for an additional category
of benefit – to obtain an in-depth understanding of firms’ organizational culture and practices,
especially as they might affect a prospective female employee. Consequently, we suggest, such
“differentiated networking” requires women to network more than men do.
“SCOUTING” FOR GOOD JOBS
The #MeToo movement – which emerged on social media in the fall of 2017 to call
attention to the pervasiveness of sexual misconduct, harassment and assault in the workplace and
elsewhere – brought to light the widespread existence of “whisper networks.” Whisper networks
refer to informal conversations between women, often those in male-dominated fields, about
which men to “watch out for,” how to avoid harassment and, more broadly, how to survive and
thrive in the workplace. Whisper networks are reputed to discuss workplace sexual misconduct,
harassment and assault, but also more mundane areas in which women might encounter gender-
based obstacles, like how to handle office politics, salary negotiations, promotion processes and
work-life balance issues. Recent journalistic reporting in the wake of the #MeToo movement has
10
revealed many anecdotes of women in media, technology and finance who relied on such
networks,4 but little systematic research has examined such information sharing among
professional women.
The #MeToo movement crystallizes three key insights for research on gender, networks,
and labor markets. First, scholars have long known that women’s success in organizations is
shaped by the organizational environment. While the #MeToo movement has highlighted how
abuses of power and sexual harassment shaped women’s careers, it also pointed to a broader set
of issues that impact women at work. #MeToo revealed that not only behavior of men in
positions of power, but broad organizational culture and practices shape women’s opportunities
to thrive in a particular organization. Research has long emphasized that organizational
characteristics of the firm a woman joins – such as organizational culture, demography, and HR
policies – will have important consequences for her future career trajectory (Kanter 1977; Ely
1994; Kalev, Dobbin and Kelly 2006; Turco 2010). More recently, Groysberg (2010) suggested
that for precisely this reason, women may be more careful in assessing prospective employers
than men are.
Second, it is not only researchers who know this; professional women know it too. As a
result, they often seek advice from other women on how to navigate these realities. Research has
long indicated that women use gender-based networks to receive support, advice, and other
resources they can’t receive from men (Lincoln and Miller 1979; Lockwood 2006). Mostly
notably, Ibarra (1992) showed that women have functionally differentiated networks, seeking
social and emotional support from other women, but instrumental and task-based support more
from men. But the #MeToo movement revealed that gender-based networks not only provide
4 See https://www.nytimes.com/2017/11/04/business/sexual-harassment-whisper-network.html.
11
friendship and emotional assistance, but also particular types of information about the job, the
bosses, or the firm more generally, that can help women survive and thrive professionally. Thus,
we make an argument that builds on and extends Ibarra’s (1992) “functionally differentiated
networks” argument: that in addition to social and emotional support, women seek important
gender-specific information and advice from other women on how to navigate their careers as
women and that this process begins during the job search. To do so, they engage in
“differentiated networking”.
The third insight – and our primary theoretical contribution – is that women use
information provided by these networks to “scout for good jobs”: that is, to gain an in-depth
understanding of the organizational culture and practices, especially as they might affect female
employees, at target firms. Prior research has documented that women’s expectations about
gender-based obstacles in particular firms or occupations shapes their job search strategies. In
particular, studies highlight that women engage in pre-emptive avoidance of discrimination: they
steer themselves out of applying for jobs where they expect to encounter gender-based barriers.
For example, Barbulescu and Bidwell (2013) reveal that because female MBAs believe that they
will be less likely to receive job offers in finance, a field historically inhospitable to women, they
chose not to apply to these jobs. Other research shows that women steer themselves out of
applying for jobs in technology (Fernandez and Friedrich 2011) or management (Storvik and
Schone 2008; Fernandez and Campero 2016). Others show that not only women steer themselves
out of applying for jobs in certain occupations, they also adjust their search strategy in other
ways as well. For example, Pager and Pedulla (2015) show that women apply to a narrow range
of occupational categories than do men. And Sterling (2017) finds that because internships offer
disadvantaged candidates a chance to overcome statistical discrimination by revealing their
12
ability, women prefer to enter the firm through an internship rather than through direct
application for a job.
While little research has investigated how women form perceptions of anticipated
gender-based barriers in particular firms or occupations, it is plausible gender-based networks
are an important part of this process. Some qualitative evidence suggests that women network
with other women to obtain information about gender-based obstacles at particular employers, or
to “scout for good jobs”. Most notably, in a study of engineers in Silicon Valley technology
firms, Shih (2006) reports that in Silicon Valley tech firms, white women actively used their
networks to learn about which companies are more egalitarian and “women-friendly”. She
writes: “These women did not simply switch from company to company in the hopes of finding a
less biased environment but rather engaged in a careful scrutiny of potential companies, which
includes soliciting information and advice about which companies are more egalitarian and
evaluating the prospective new companies”(p. 186). She notes that women engineers reported
“seeking and offering “backstage” information about which places or people to avoid, how to
deal with projects, and how to find other job opportunities. This ethic is rooted in the experiences
of women as minorities in engineering and in their belief that they share similar obstacles” (p.
196). Shih also found a similar strategy among Asian engineers who used ethnic networks to find
less biased employers or to enter entrepreneurship.
Recent reporting on #MeToo has suggested that networks described by Shih (2006)
appear to still be relevant in technology firms. For example, a recent New York Times article5
reports that invitation-only Facebook groups, such as Tech Ladies, share information about job
openings and networking events for women working in technology companies and also has a
5 See https://www.nytimes.com/2017/11/04/business/sexual-harassment-whisper-network.html.
13
forum where members can learn about other women’s experiences with an employer’s work
culture and practices, including issues of sexual harassment. Allison Esposito, the founder of
Tech Ladies, explains that women use the forum to ask for advice when applying for new jobs,
telling the New York Times: “What we’ll see is someone post that they are applying for a job at
a company and ask if anyone has any good or bad stories about it … If it was great, you’ll see
that in the comments. But if it wasn’t, people will say ‘DM me’ or ‘let’s take it to the phone,’ to
share the information.”
While #MeToo movement brought to the popular attention the existence of these
“whisper networks”, the evidence for them in the scholarly literature is limited. In this paper, by
examining differences in digital traces of networking outreach by a population of young, high-
potential MBA students, we provide important new empirical evidence for this phenomenon.
EMPIRICAL SETTING
The key empirical challenge for studying gender differences in network outreach in a
field-based setting is to identify outreach behavior independent of how it was received. We
attempt to do this by leveraging a unique dataset on the use of the alumni database by job-
seeking students in an elite MBA program. Most notably, our data contains digital traces of
students’ outreach to alums. Thus, we do not restrict our sample to successful networking
attempts (as, for example, would a study focusing on the addition of LinkedIn contacts), which
would potentially bias results by excluding those interactions where the networking attempt was
ignored or rejected.
It is also important to note a few additional features of our setting that make it attractive
for our study. Focusing on MBA students at a single university – a relatively homogenous group
– enables us to examine networking behavior among comparably qualified women and men,
14
ruling out some potential confounders such as quality of education, prior experience and other
types of human capital. Such variation exists, of course, but within a restricted range; and more
importantly, alumni rarely look at differences in human capital, relying instead on the screening
of the school’s admissions office. In addition, familial responsibilities, another potential factor
that might restrict women’s opportunities to network (Forret and Dougherty 2004), are relatively
scarce in our setting, as most MBA students are young professionals (median age is 30) and over
94% are without children. Finally, nearly all MBA students engage in the search for an
internship. This allows us to study networking behavior without concern about self-selection into
job search (though we acknowledge that not all job searches involve equal amounts of
networking).
Networking activities are critical in MBA job search, as students use contacts to identify
possible career opportunities, learn about prospective employers, prepare for interviews, and find
advocates and mentors within hiring organizations (Sterling 2014; Barbulescu 2015).
Furthermore, alums are particularly useful in helping students to determine fit with prospective
employers (Greenberg and Fernandez 2016). While networking with alums is certainly not the
only type of networking MBA students engage in, access to an alumni network is an important
selling point for many MBA programs, including the one we studied. We chose our research
setting because the alums of this institution are known to be remarkably responsive to its current
students. Indeed, the leading aggregator of rankings and other resources about business schools
has evaluated our research site as having the best alumni network in the world, calling its
graduates “the perfect alums to approach when you’re looking for guidance.” Anecdotally,
students report that alums almost always get back to them, often within an hour or two.
15
Our setting also rules out an alternative explanation for gender homophily in networking
as arising from differential access to contacts. In other settings, people often access new contacts
through referrals from their existing contacts. If women tend to have more female contacts than
men do, the referral mechanism will result in women having more opportunities to interact with
women than men do. This explanation is largely ruled out in our setting in two ways. Most
importantly, the university gives all students access to an online alum database, thus, providing
them with a similar pool of potential contacts. In essence, the university itself acts as referrer,
giving all students equal access to all alumni. This database contains information about all living
alums of the school, including their name, class year, gender, ethnicity, citizenship, prior
education, contact information, and information about their current and some prior employment,
including firm name, job title, industry, and job function. Some records, particularly those for
recent alums, include a photo. The database is searchable by any of the fields above. Most
information is quite current, with the median record updated just 18 months prior to the start of
the academic year that we studied.
Further, we do not believe that women use the online alum database to reach out to
female alums because they find it more difficult to meet women through other means. To start,
we find little evidence that students use the online database in lieu of face-to-face networking.
Women and men in our study were equally likely to attend on-campus recruiting events, and
when we include the number of company presentations attended as a covariate in our models, we
find that attendance at company briefings has a weak, positive association with networking via
the alums database. And we have reasons to believe that in face-to-face networking female
MBAs have many opportunities to interact with other women. For example, Rivera documents
that the employers consciously strive to send to campus diverse recruiting teams (Rivera 2015:
16
70). This means that women frequently have access to female contacts during recruiting visits,
and are not reaching out to women online in response to the lack of diversity in on-campus
recruiting. This makes it less likely that women turn to online interaction with alumni primarily
because they are unable to reach out to women through other means.
Finally, we note that women’s networking outreach was not influenced by university
efforts to promote gender-based networking. To start, most networking advice to women in the
popular and practitioner press stresses the importance of women overcoming their reputed
preference to network with women, in order to network more with men and high-status actors
(e.g., CITE)6. Journalistic reports describe whisper networks as a response to discrimination, not
as a proactive strategy to avoid discrimination. In our setting, the students we interviewed know
of no resources that provide networking guidance, help, or advice targeted at one gender group
or another. We specifically asked four female students at how they arrived at their differentiated
networking strategy; they all emphasized being informed by their personal experience and not by
the university’s career development office.
RESEARCH DESIGN AND METHODS
We examine how a complete cohort of 287 first-year MBA students used the alumni
database in their searches for summer internships. As in prior work on MBA job search (e.g.,
Sterling 2014), we analyze internship searches by first-year students, rather than searches for
full-time employment by second-year students, for greater comprehensiveness and temporal
synchrony. In business, as in other professional occupations, employers often use internships to
assess a person’s abilities over a period of weeks or months and then determine whether or not to
6 https://www.forbes.com/sites/samanthaettus/2017/06/15/why-women-arent-networking/#c0007f2aeadb
17
offer her a job. In the two-year MBA program where the data for this study were collected,
virtually all first-year students search for a summer internship, which they consider to be a
critical step in the search for full-time employment after graduation. Because employers make
offers to some students following the internship, many students do not search for jobs during the
second year; this fact would create a sample selection issue for studying the search for full-time
jobs by second-year students. Further, the timing of the full-time job search varies more widely
across students, compared to the search for first-year internships, which is relatively compressed.
For these reasons, we focus on the search for first-year internships, rather than post-graduation
jobs.
This research required the collection of three distinct data sets.7 First, and most notably,
we collected server logs of students’ use of the alumni database. Students using the database can
search alums’ profiles using keywords, industry tags, firm or person names, and a variety of
other means. Logs of which alums appeared in their search results were recorded over an eleven-
month period, beginning in the summer prior to the matriculation of first-year students, when
they first gained access to the database, and through the end of their first year, when virtually all
students had started their internships. In particular, we logged each time a student viewed an
alum’s profile page and each click on the “mailto:” link (an “emailclick”), an action that initiates
a new email from the student to the alum. For each such emailclick, a precise timestamp and the
ID numbers of the searching student and the target alum were logged. Thus, rather than relying
on self-reports of past networking, we track actual emailclick behaviors, coming as close as
possible to observing networking (albeit of one particular type) directly. Anecdotal accounts
7 All of these data sets are linked through the use of anonymous identifiers, which enable us to link the data about
individual students and alums with the database activity logs while protecting the privacy of both students and
alums.
18
suggest that an emailclick from the alum database is by far the primary means by which students
initiate contact with alums.
Second, we collected individual-level data about all living alums of the school. The alum
data included each alum’s gender, employer, industry, and job title and description. The job
description data were selected by alums themselves from a typology of thirteen possible titles,
ranging from “Analyst/Associate/Consultant” and “Student/Intern/Resident” to
“Partner/Principal/Managing Director/VP” and “CEO/President/Chairman.” Using this
information, we identified senior-level alums as those with titles of “CEO/President/Chairman,”
“CxO,” “Partner/Principal/Managing Director/VP,” “Member of Board of Directors,” and
“Owner/Founder” and junior-level alums as those with any other titles. Overall, the alum
population is 78.2% male and 21.8% female. If we restrict the alumni population to those who
graduated in the prior twenty, thirty or forty years, the proportion of female alums rises to 30.8%,
29.7% and 28.0%, respectively. Also, female alums are not confined to particular jobs or
industries: if we restrict the alumni population to those in job functions or industries in which
many students seek employment, the gender composition remains substantively similar.
Third, we assembled individual-level data on the students from three sources. The
registrar provided data on each student’s gender, citizenship, native language, and ethnicity;
campus residence status (i.e., whether they lived on campus or off); class section; relative
GMAT score8; and marital status. The career development office provided data on students’
attendance at company briefings as well as data from two student surveys. The pre-matriculation
survey conducted in August inquired about each student’s intentions regarding the firms,
industries, job functions, and geographic locations in which they planned to search for
8 For reasons of confidentiality, GMAT data were provided to us not as raw scores, but as standardized variables,
calculated relative to this cohort of students, with a mean of zero and a standard deviation of one by construction.
19
internships. We exclude from our sample ten students who indicated that they did not intend to
search for internships, mostly because they were pursuing dual degrees (primarily MD/MBA
students) or would be returning to a previous employer9. The internship outcome survey
conducted in May collected information on students’ self-reported satisfaction with the
internship received; note that this survey was conducted after the internship offer was signed, but
before the internship began. We also added some items to the May survey to evaluate the validity
of our emailclick measures. Lastly, we conducted our own survey in October, collecting
psychometric data and data on students’ networking strategies.
Variables
Our primary dependent variable, emailclick, is a count variable equal to the total number
of clicks on alums “mailto:” links made by each student. We argue – and below present some
evidence to support this assertion – that this is an excellent proxy for the number of emails a
student initiated to alums with whom they were not previously acquainted. To examine the
gender distribution of each student’s networking targets, we split the count of alums contacted
into subsets of female (emailclick_f) and male (emailclick_m) alums; by construction, emailclick
= emailclick_f + emailclick_m. To examine the hierarchical status diversity of each student’s
networking targets, we split the count of alums contacted into those who have VP or higher
status in their organizations (emailclick_vp) and those who are more junior (emailclick_jr); by
construction, emailclick = emailclick_vp + emailclick_jr.
Our main explanatory variable is Female, coded as one for female students and as zero
for male students. We also create a number of control variables that are likely to affect
9 Note, however, that some dual-degree and sponsored students indicated that they nevertheless did intend to search
for internships; these were retained in our primary sample.
20
networking and on which women and men might plausibly differ. We expect that the personality
trait Extraversion is associated with intensity of networking behaviors (Forret and Dougherty
2001; Shipilov et al. 2014) and may be associated with gender (Lynn and Martin 1997 show a
correlation in the general population, but Feiler and Kleinbaum 2015 find no correlation in an
MBA student sample), so we measured it using the extraversion scale in the Big Five Inventory
(John and Srivastava 1999). We expected that students who are less occupationally focused
might search more broadly, and that occupational focus might co-vary with gender (Barbulescu
and Bidwell 2013), so using information from the October career survey, we created a variable
Search Breadth to measure the number of job functions in which a student expressed an interest
in working.10
We controlled for demographic characteristics. We created three dummy variables
(Asian, White and Other) to control for students’ race and another dummy variable to indicate
whether the student is a Native English Speaker. We created a dummy variable Sponsored for
those students whose tuition was paid by their past employer, in return for a promise that they
would resume their employment after business school. Because these students are likely to return
to their employer upon graduation, their job search motivations – and, consequently, their
networking patterns – might differ from those students who are not sponsored (though in many
cases, sponsored students still search for a summer internship with a different employer). Lastly,
we include two measures of human capital. A continuous variable GMAT (std) measures the
distance in standard deviations between a student’s score on the Graduate Management
10 We have also re-estimated our models with controls for the number of firms and number of industries students
expressed an interest in. Across all preliminary models, these variables were not significant and their inclusion did
not substantively affect our results. As a more behavioral indicator of search breadth, we substituted the number of
company briefings attended in place of the Search Breadth variable defined by job functions. Again, the results were
substantively unchanged. Finally, we dropped this control variable altogether to see whether another covariate
(especially gender) would pick up this variation; these results were also substantively unchanged.
21
Admissions Test and the mean GMAT score in the sample. And Log Work Experience is the
natural logarithm of the number of years of professional experience prior to beginning business
school. We calculated years of work experience as the number of years between the end of the
student’s undergraduate degree and the start of business school, less the number of years spent in
other educational programs, as indicated in students’ reporting to the registrar.
RESULTS
We begin by presenting some descriptive statistics. Means, standard deviations and inter-
correlations for all variables are presented in Table 1. To support the validity of emailclick as a
behavioral indicator of networking activity, we note that students searching for jobs had more
emailclicks than those not searching for jobs (p < 0.04) and those who reported in our October
survey that they viewed the alum database as a valuable job search resource had more
emailclicks than those who did not (p < 0.001). To further assess the validity and reliability of
emailclicks as a measure of networking activity, we included in the May survey a page in which
we showed respondents the names and employers of some alums whom server logs indicate they
had previously emailclicked and other alums whom server logs indicate they did not emailclick.
We then asked them about their interactions with these alums. We found that when we observed
an emailclick to a specific alum, students reported having interacted with that alum 78% of the
time; conversely, when we observed no emailclick, students reported having interacted with the
alum only 16% of the time11. Given imperfect recall in survey response and other channels of
11 For greater comparability, and recognizing that the majority of alums have no interactions with students at all, the
non-emailclicked alums were selected from among those with the highest rates of interaction with students other
than the focal student.
22
possible interaction between students and alums, we found these results to be strong evidence
supporting the validity of emailclicks.
<Insert Table 1 about here>
Overall, we find little evidence of gender differences in whether or not students used the
database. The distribution of emailclicks across students is skewed12: about one-third (35.3%) of
students did not emailclick any alums. Selection models (available upon request) indicate that
gender is not a significant predictor of positive use of the alum database (p = 0.796), so given
equal opportunity to network with alumni, women take advantage of that opportunity at a rate
equal to that of men. Also, no other key covariate predicts which students choose not to
emailclick any alums. Indeed, the only significant predictors of positive (versus zero) emailclicks
are a stated interest in jobs in financial services (p = 0.023) or human resources (p = 0.012).
Anecdotally, students report a belief that financial services is an industry in which networking
beyond the formal recruiting process is de rigeur; conversely, relatively few HR positions are
available through on-campus recruiting, so networking with alums may be an alternative avenue
to finding such a job.
Our descriptive results would seem to confirm the intuition that the use of the database is
closely linked to internship search activity, especially to early stages of learning about job
opportunities, identifying potential employers, and networking with employees at firms of
interest. Emailclicks occur disproportionally before internship offers are received (84.2% for
men; 85.3% for women; p = 0.321); for comparison, the median internship offer was received on
12 But our robustness section below indicates that our results are not driven by outliers.
23
February 10th, 59.5% of the way through the academic year13. We also find that both female and
male students use the database to network broadly: 95% of emailclicks are targeted at alums who
do not work at the firm where the student ended up interning. This result also does not differ
significantly by gender, whether we look at all networking activity throughout the year (p =
0.363) or only networking activity occurring before the student received an internship offer (p =
0.471).
Before moving on to multivariate analyses, we descriptively examined networking
behaviors, focusing on differences between male and female students, in Table 2. The most
striking descriptive result is that women, on average, reach out to fully 63% more alums (6.5 vs.
3.99; p = 0.031) than men do. Further, this difference appears to be explained by the facts that
women, compared with men, contact nearly three times more female alums (2.42 vs. 0.88; p <
0.001) and at least as many male alums (4.08 vs. 3.10; p = 0.14); and that women contact nearly
twice as many junior alums (5.31 vs. 2.96; p < 0.02) and at least as many senior-level alums
(1.19 vs. 1.03; p = 0.27).
Comparing these descriptive statistics to the alumni population as a whole, we note that
for male students, the aggregate gender distribution of alums contacts closely parallels the gender
distribution of the alumni population: collectively, 77.8% of male students’ emailclicks were to
male alums and 22.1% were to female alums, a distribution that is indistinguishable from the
gender distribution of the alum population (p > 0.4). Female students, however, directed 37.2%
of their emailclicks to female alums, a rate significantly higher (p < 0.0001) than women’s
representation in the alums population. In contrast, senior-level alums (24.5% of the database)
were overrepresented among the contacts of male students (28.3% of emailclicks; p < 0.02), but
13 The median male student received an offer on February 11th; the median female student received an offer on
February 9th.
24
under-represented among the contacts of female students (19.9%; p < 0.01), relative to their
representation in the database; this univariate result is inextricably linked to women’s greater
propensity to contact female alums and points to the need for multivariate analysis.
<Insert Table 2 about here>
We present the results of multivariate Poisson quasi-maximum likelihood regressions in
Table 3, beginning with control variables14. Poisson count models are in the linear exponential
family, so the conditional mean of the data is assumed to be correctly specified, but no additional
distributional assumption is required to generate consistent coefficient estimates (Silva and
Tenreyro, 2006). We find little difference in networking behavior by students of different
ethnicities: relative to their white peers, the Asian coefficient is statistically insignificant in all
models. Students of Other Ethnicities (other than white or Asian) may use the alum database
less, particularly to emailclick senior-level alums. Non-native English speakers may also use the
alum database less, but again, the effects are inconsistent. Students engaging in a broad job
search (that is, those who indicated interest in more job functions on our October survey) tend to
use the alum database more and students whose tuition was sponsored by a previous employer –
to whom they are committed to returning – use it less. People with extraverted personalities use
the alum database more, but the effect is only estimated precisely enough for statistical
significance in interactions with female alums.
<Insert Table 3 about here>
We test for the presence of gender differences the number of contacts a student reached
out to in Model 1, where the dependent variable is the total count of emailclicks. Controlling for
other observable demographic characteristics, we find that female students, on average, click on
14 Because U.S. Citizenship is highly correlated (0.71) with Native English Speaker, we drop the citizenship variable
to avoid problems of multicollinearity. Results are substantively unchanged if we retain citizenship instead.
25
the mailto links of 43% more alums than their male classmates (exp[0.360] = 1.43; p < 0.05). To
examine the role of contacts’ gender in differences in students’ networking, we look to Models 2
and 3, whose dependent variables are emailclick_f, the count of emailclicks directed to female
alums, and emailclick_m, the count of emailclicks directed to male alums, respectively. In Model
2, female students mobilize ties to female alums at 2.27 times the rate of their male classmates (=
exp[0.819]; p < 0.001). To better contextualize this result, we note that female students do not
reach out to female alums at the expense of ties to male alums: Model 3 indicates that women
may also reach out to more male alums, though the effect size is modest and imprecisely
estimated (exp[0.174] = 1.19; p = 0.350). To examine the role of contacts’ organizational status
in students’ networking, we look to Models 4 and 5, whose dependent variables are
emailclick_jr, the count of emailclicks directed to junior-level (i.e., below the vice-president
level) alums, and emailclick_vp, the count of emailclicks directed to senior-level alums,
respectively. Female students network with junior-level alums at a rate 56% greater than that of
their male classmates (exp[0.444] = 1.56; p < 0.05). Again, to better contextualize this result, we
note that female students do not mobilize ties to junior-level alums at the expense of ties to
senior-level alums: Model 5 indicates that women mobilize ties to senior level alums at
essentially the same rate as their male classmates (exp[0.0557] = 1.056; p > 0.8).15
We replicated the analysis in Table 3 using industry controls based on students’ stated
internship search interests (Appendix, Table A1) and, alternatively, based on the industry in
which the student accepted an internship (Appendix, Table A2). In these models with industry
15 Supplemental analyses split emailclick by both alums’ gender and level at the same time. We do not include these
results because the small sample sizes resulted in imprecise estimates, but regressions indicate that women mobilize
ties to senior and junior women at approximately the same rate. Women also mobilize ties to junior men at a rate
30% higher (but with p = .193) than their male classmates and to senior men at a rate indistinguishable from their
male classmates (p > 0.6).
26
controls, we find that the Emailclick coefficient on female is positive, though no longer
significant. This could suggest that women’s greater propensity to network that we find is at least
in part driven by women seeking jobs in industries where networking is more prevalent, such as
consulting or human resources. Nevertheless, consistent with “scouting,” in all these models with
industry controls we still find that compared to men, women network with no fewer male alums
(i.e., βEmailclick_m cannot be discerned from 0), but also reach out to more female alums (i.e.,
βEmailclick_f > 0; p < 0.05). And while in both sets of models, women network with no fewer
senior-level contacts than men do, in the second set of models we find that women also reach out
more to junior alums than men do.
Due to the skewness of the emailclick variable, we worried that the effect might be driven
by a few uninhibited outliers engaging in extensive networking with alums. As a robustness
check, we replicated Table 3 using a subsample of students that excludes those who contacted at
least twenty alums. In these analyses (available upon request) the gender difference in
emailclick, emailclick_f, and emailclick_jr diminishes slightly in magnitude, but is otherwise
substantially similar, suggesting that these are not outlier effects.
Thus, to summarize the core result of this paper, when a group of MBA students are
presented with the same pool of potential contacts for networking, women reach out to male
alums at a rate that is comparable with that of their male counterparts; in addition, they
consistently reach out to significantly more female alums. We also find that while women reach
out to senior-alums at rate comparable to men, they also reach out to more junior-level alums.
Overall, this pattern means that compared to men, women spend more time and energy
networking.
27
ADDITIONAL EVIDENCE ON THE MECHANISM
We argued that our results are consistent with “scouting”, by which women network –
especially with other women – to “scout” potential employers, or to gain an in-depth
understanding of firms’ organizational culture and practices, especially as they might affect a
prospective female employee. We recognize, however, that the literature suggests a number of
alternative explanations for why women might network with other women. We have discussed
how the explanations that focus on availability of contacts do not apply in our setting. In this
section, we examine another important potential explanation – that female alumni are more likely
to reciprocate women’s networking outreach. We conclude by discussing our interview evidence
which illuminates specific resources female MBAs get from networking with female alums.
Women network more with women because they get less help from male alums. We
argued that because all students can use the alum database, they have access to the same pool of
potential contacts. However, as we noted some research suggests that women might receive less
help from men (e.g. McDonald 2011). This might potentially lead women to network more with
women.
We empirically investigated these issues using our emailclick validation survey.16
Recall that in the survey, conducted in May, students were presented with a list of alums and
asked about their interactions with each alum and what type of help (if any) the alum provided.
To examine whether female and male alums are differentially responsive to female and male
16 We also considered an alternative strategy of estimating the gender differences in benefits from the use of the
database – for example, we could examine the relationship between the number of alums contacted and job market
outcome, such as salary. However, such models would be difficult to interpret due to endogeneity: people who have
difficulties finding a job may rely more heavily than others on social networks (Loury 2006). Because our study was
designed to explore differences in networking strategy, it is not designed to estimate the causal effect of networking
on labor market outcomes.
28
students we focus on those alums whom the student reported as having personally contacted
(“reached out to through email, phone and other means”) and exclude those whom the student
met at recruiting events or through other means. Then we estimated logistic regressions,
clustered by student, of two proxies for non-response to the students’ networking attempts:
students’ reports that a) they did not interact with this alum in spite of having reached out or that
b) they did not receive help from this alum. We find no statistically significant differences by
alum gender, suggesting that alums show no gender preference in their propensity to respond to
students’ outreach.
To examine whether female and male alums are differentially helpful to female and male
students, we focus on those alums with whom the students reported interacting. When an
interaction was reported, a follow-up question prompted the student to indicate what kind of help
the alum provided from a list. While some types of networking help listed (e.g. “provided
general career advice,” “informed me about a job opening”) might only require alums’ time and
effort, others might be more potentially socially costly (e.g. “introduced me to his/her contacts,
“put in a good word for me at his/her company”, “acted as a formal referrer for me”). Alums
might consider how providing this type of help might impact their own reputation with their
employers or professional contacts (Smith 2005). Because research finds women might be
disadvantaged in receiving more socially-costly types of help (McDonald 2011; Abraham 2017),
we coded these outcomes using three dummy variables, representing different levels of help: Any
Help was coded as 1 if the student reported receiving any type of help; Costly Help was coded as
1 if the student received an “introduction,” a “good word,” or a “referral”17; and Referral was
coded as 1 if the student received a specific job referral. Then, using data on all the alums that
17 Our interview with the school’s director of career development indicated that an “introduction,” a “good word,” or
a “referral” were the types of help that are most significant and most valuable to students.
29
the student contacted, we estimated dyad-level, random-effects linear probability models of
helping behavior with errors clustered on the student. These analyses (see Appendix, Table A3)
indicate that the gender of the alum has no statistically significant effect on the likelihood of
receiving help or the amount of help received for female and male students.
Taken together these results suggest no gender difference in alums’ helpfulness, at least
as reported by students: conditional on a student reaching out to an alum, both female and male
alums appear to be equally likely to respond to female and male students and to provide them
with the same amount of help. We conclude that the incremental networking that female job-
seekers engage in compared to their male peers is not driven by any gender difference in the
helpfulness of contacts. So why do women network more, especially with other women?
Differentiated networking and the benefits of scouting. In an intra-organizational
setting, Ibarra (1992) argued that women have differentiated networks, consisting of women
whom they turn to for social and emotional support and men whom they rely on for instrumental
access. In the context of job search, we suggest that women engage in “functionally
differentiated networking,” reaching out to women to “scout for good jobs.” That is, there are
certain types of information and advice that women value more highly than men do and that they
receive primarily from other women. In particular, because women anticipate encountering
gender-based obstacles in the workplace, they seek out information about how to navigate those
obstacles from other women.
Some interview evidence here
30
DISCUSSION
Although networking is a key to professional success, we know little about gender
differences in networking behavior. We argued that women use network outreach to other
women to “scout for good jobs,” or to identify employers where they have a greatest chance of
professional success. Because they reach out to more women than men do, women end up
networking with more people. Our empirical results confirmed that in this setting MBA students,
while reaching out to same number of male alums as men as men do, also reach out to more
female alums. We also tested and found little evidence for an explanation for why women might
network more in our setting than they otherwise would: that women network more because they
get less help from alums. Echoing the “whisper networks” that came to light during the #MeToo
movement, our interview evidence suggests that female MBA students sought information from
female alums about issues that might impact their professional lives as women, such as parental
leave, work-life balance issues, biases in promotion opportunities and representation of women
in management positions.
Like all research, this study is not without limitations. An important advantage of our
setting is that we systematically examined networking outreach behavior within a well-defined
population of students and alumni. Although this research strategy enabled us to observe
networking outreach more comprehensively than in prior research and was unbiased by the
success of the outreach attempt, our digital trace data did not allow us to examine the content of
subsequent communications across any of the various media that students used to interact with
alumni. Thus, although we have strong evidence that women reach out to women at a higher rate
than men do, we do not have direct quantitative evidence that scouting is the reason why. To
31
provide preliminary evidence for this mechanism, we conducted a series of interviews, which
provided qualitative support for our argument. We also buttress our claims with journalists’
accounts of “whisper networks” through which women scout out bosses and firms. We leave it to
future research to assess how broadly representative these results are.
Another limit on the generalizability of our results stems from features of the MBA
setting itself. Our study is conditional on its population of students in an elite MBA program.
While this idiosyncratic setting offers the benefits that all students are, to a first approximation,
comparable in their human capital and all have access to the same population of alums, these
benefits come at a significant cost. It goes without saying that this population is selected from the
broader population in a decidedly non-random way. And while prior research (Shih 2006) and
the journalistic accounts of “whisper networks” revealed by the #MeToo movement suggest that
our findings are generalizable to industries where women are a negatively stereotyped minority,
more research is needed to understand gender differences in networking in other parts of the
labor market.
In spite of these limitations, we believe this study makes a significant contribution to the
literatures on networks and gender. Our first contribution is to propose a novel theoretical
mechanism “scouting” to explain gender differences in networking strategies we observed. In
this way our study builds upon and extends the notion of “functionally differentiated networks”
(Ibarra 1992, 1997) in two ways. Most importantly, our interviews suggest that women engage in
“differentiated networking.” That is, they use ties to women are not only to gain friendship and
emotional support, but also to exchange important gender-specific resources. We argued that
through relationships with other women, women gather information that is critical in helping
them make better career decisions. Second, our study also suggests that gender homophily in
32
women’s networks is not simply a result of men blocking women’s networking attempts (c.f.
Mehra, Kilduff and Brass 1998). Consistent with other studies that describe the benefits of same-
gender relationships for women (Lincoln and Miller 1979; Lockwood 2006), we show that
during networking women seek out other women, because they derive unique benefits from these
relationships.
While our study suggests that women might network successfully to find jobs where they
have a fair chances of professional success, it is important to note that these “differentiated
networking” strategies might come at a cost. The most obvious cost in our context is the time and
effort required in networking, an activity that many women and men alike consider not highly
enjoyable or even outright unpleasant (Casciaro, Gino and Kouchaki 2014; Kuwabara,
Hildebrand and Zou 2016). Because time is the scarcest resource of all busy professionals, the
incremental networking that women do comes at the expense of other career-enhancing
activities. In addition, as Ibarra (1992) shows “functionally differentiated networks” may result
in in lowered network centrality. Lowered centrality might make women appear as less desirable
protégés or exchange partners, further constraining their ability to develop networks. As such,
“scouting” in the job search constitutes a kind of discrimination insurance that is costly to
women, but which, they hope, will prevent larger challenges later.
The idea of “scouting” also helps to theoretically situate recent discussions prompted by
the #MeToo movement of “whisper networks” through which women share information about
men to watch out for. Our results suggest that these “whisper networks” are not an isolated
phenomenon related to sexual misconduct, harassment and assault, but rather one – perhaps the
most extreme – manifestation of women’s attempts to “scout out good jobs” by seeking and
sharing information with other women. Doing so is a rational response to the possibility of
33
encountering gender-based obstacles in the workplace, and, helps women to avoid the kinds of
work environments where they will face the biggest constraints on their success and
advancement (Shih 2006). And because we suggest that gender differences in networking are not
an outcome of innate gender differences, but rather a rational response to the existence of
gender-based obstacles especially in male-dominated industries, we think our model likely
applies to other historically disadvantaged groups (c.f. Shih (2006) on Asian engineers in
technology firms)
Our second theoretical contribution is to bringing to light gendered processes of tie
formation and mobilization. Despite the evidence that the benefits individuals derive from
network do not explain how they create and mobilize social ties (Ryall and Sorenson 2007;
Buskens and van de Rijt 2008; Obukhova and Lan 2013, Kwon and Adler 2014), most research
in the social network field has focused on network structure, largely neglecting the agentic
processes of tie formation and mobilization (Casciaro, Gino and Lobo 2014: Kuwabara,
Hildebrand and Zou 2016). The neglect of attention to these processes is problematic for future
development of the social network field, as it tries to understand not only career consequences of
individual differences in networks, but also processes that give rise to these differences in the
first place (Kilduff, Tsai and Hanke 2006; Ahuja, Soda and Zaheer 2012; Burt 2012). By
isolating the role of gender in giving rise to individual differences in networking outreach, our
study makes an important step toward revealing processes that lead to individual difference in
networks. And by revealing differences in how individuals network our study brings us one step
closer to understanding the role of agency in social networks.
Our theoretical contributions have significant implications for organizational and societal
efforts to promote gender equity by highlighting potential benefits of providing women with
34
greater networking opportunities, as exemplified in our case with the access to the alum
database. While many employers have realized the importance of using diverse recruiting teams
to provide job candidates with opportunities to interact with someone like them (Lockwood
2006) to learn informally about the employer (Rivera 2015: 45, 70), our study highlights that
contacts can also come from alum databases. This is important as increasingly, it is not only
universities, but also firms who think about their former affiliates as “alums” (e.g., The
Economist 2014; Forbes 2016). As internet technologies make such databases of potential
contacts more accessible than ever before, more and more organizations are investing in them.
And our results suggest that these institutional efforts to create networking opportunities are a
valuable, if costly, response to gender-based barriers in the workplace, in that they give women
greater ability to “scout out” potential employers to identify those where they will have a fair
chance to succeed.
35
REFERENCES
Abraham M. 2017. Explaining unequal returns to social capital among entrepreneurs.
Unpublished manuscript. Columbia University.
Ahuja G, Soda G, Zaheer A. 2012. The Genesis and Dynamics of Organizational Networks.
Organization Sci. 23(2): 434-448.
Barbulescu R. 2015. The strength of many kinds of ties: Unpacking the role of social contacts
across stages of the job search process. Organ Sci. 26(4) 1040-1058.
Barbulescu R, Bidwell M. 2013. Do Women Choose Different Jobs from Men? Mechanisms of
Application Segregation in the Market for Managerial Workers. Organ Sci. 24(3) 737-756.
Brands, R.A., I. Fernandez-Mateo. 2016. Leaning out : How negative recruitment experiences
shape women's decisions to compete for executive roles. Administrative Science Quarterly 62(3)
405-442.
Brands RA, Kilduff M. 2014. Just like a woman? Effects of gender-biased perceptions of
friendship network brokerage on attributions and performance. Organ Sci. 25(5) 1530-
1548.
Burt RS 2012. Network-Related Personality and the Agency Question: Multirole Evidence from
a Virtual World. Am. J. Sociol. 118(3) 543-591.
Casciaro T, Gino F, Kouchaki M. 2014. The contaminating effects of building instrumental ties:
How networking can make us feel dirty. Administrative Science Quarterly. 59(4) 705-735.
Castilla, E.J., G.J. Lan, B.A. Rissing. 2013. Social networks and employment: Mechanisms (part
1). Sociology Compass 7(12) 999-1012.
Eagly AH, Karau SJ. 2002. Role congruity theory of prejudice toward female leaders. Psychol.
Rev. 109(3) 573-598.
The Economist. 2014 “Corporate Alumni: Gone But Not Forgotten.” (March 1, 2014). Last
accessed June 21, 2017 at: http://www.economist.com/news/business/21597935-more-firms-are-
seeking-stay-touch-former-staff-gone-not-forgotten.
Ely RJ. 1994. The effects of organizational demographics and social identity on relationships
among professional women. Administrative Science Quarterly. 39(2) 203-238.
Feiler DC, Kleinbaum AM. 2015. Popularity, Similarity, and the Network Extraversion Bias.
Psychol. Sci. 26(5) 593-603.
Fernandez, R.M., C. Friedrich. 2011. Gender Sorting at the Application Interface. Industrial
Relations: A Journal of Economy and Society 50(4) 591-609.
Fernandez-Mateo, I., R.M. Fernandez. 2016. Bending the pipeline? Executive search and gender
inequality in hiring for top management jobs. Manage. Sci. 62(12) 3636-3655.
Flynn, F.J. 2007. Paths to Power: An Interview with Heidi Roizen. J. Pfeffer, ed. Stanford GSB
Videocase OB-70V.
Forbes 2016. “Why McKinsey & Company’s Alumni Networks Is Crucial To Its Success.” Last
accessed June 21, 2017 at https://www.forbes.com/sites/davidburkus/2016/07/05/why-mckinsey-
companys-alumni-network-is-crucial-to-its-success/.
Forret ML, Dougherty TW. 2001. Correlates of networking behavior for managerial and
professional employees. Group and Organization Managementl. 26(3) 283-311.
Forret ML, Dougherty TW. 2004. Networking behaviors and career outcomes: Differences for
men and women? J. of Organizational Behavior. 25 419-437.
Greenberg J, Fernandez RM. 2016. The strength of weak ties in MBA job search: A within-
person test. Sociological Science.
36
Groysberg B. 2010. Chasing Stars: The Myth of Talent and the Portability of Performance.
Princeton University Press, Princeton, NJ.
Heilman ME, Block CJ, Martell RF, Simon MC. 1989. Has anything changed? Current
characterizations of men, women, and managers. J. Appl. Psychol. 74(6) 935-942.
Huffman M, Torres L. 2002. It's not only "who you know" that matters - Gender, personal
contacts, and job lead quality. Gend. Soc. 16(6) 793-813.
Ibarra H. 1992. Homophily and differential returns: Sex-differences in network structure and
access in an advertising firm. Administrative Science Quarterly. 37(3) 422-447.
Ibarra, H. 1997. Paving an alternative route: Gender differences in managerial networks. Social
Psychology Quarterly 60(1) 91-102.
Ingram P, Zou X. 2008. Business friendships. Emerald Group Publishing Ltd, Bingley.
John OP, Srivastava S. 1999. The big five trait taxonomy: History, measurement, and theoretical
perspectives. Pp. 102-138. Handbook of Personality: Theory and Research, Pervin LA, John OP,
ets. Guilford Press, New York.
Kanter RM. 1977. Men and women of the corporation. Basic Books, New York.
Kilduff M, Tsai W, Hanke R. 2006. A paradigm too far? A dynamic stability reconsideration of
the social network research program. Acad. Manage. Rev. 31(4) 1031-1048.
Kleinbaum AM, Stuart TE, Tushman ML. 2013. Discretion Within Constraint: Homophily and
Structure in a Formal Organization. Organ Sci. 24(5) 1316-1336.
Kuwabara K, Hildebrand C, Zou X. 2016. Lay theories of networking: How laypeople's beliefs
about networks affect their attitudes and engagement toward instrumental networking. Acad.
Manage. Rev. 2015.0076.
Kwon, S.-W., P.S. Adler. 2014. Social capital: maturation of a field of research. Acad. Manage.
Rev. 39(4) 412-422.
Loury LD. 2006. Some Contacts Are More Equal than Others: Informal Networks, Job Tenure,
and Wages. J. Labor Econ. 24(2) 299-318.
Lynn R, Martin T. 1997. Gender differences in extraversion, neuroticism, and psychoticism in 37
nations. J. Soc. Psychol. 137(3) 369-373.
Marsden, P.V., E.H. Gorman. 2001. Social networks, job changes, and recruitment. Kluwer
Academic/Plenum, New York.
McDonald S. 2011. What's in the "old boys" network? Accessing social capital in gendered and
racialized networks. Social Networks. 33(4) 317-330.
Merluzzi J. 2017. Gender and Negative Network Ties: Exploring Difficult Work Relationships
Within and Across Gender. Organ Sci. 28(4) 636-652.
Obukhova, E., G. Lan. 2013. Do job seekers benefit from contacts? A direct test with
contemporaneous searches. Manage. Sci. 59(10) 2204-2216.
Pager, D., D.S. Pedulla. 2015. Race, self-selection, and the job search process. Am. J. Sociol.
120(4) 1005-1054.
Rivera, L.A. 2015. Pedigree: How elite students get elite jobs. Princeton University Press,
Princeton, NJ.
Rubineau, B., R.M. Fernandez. 2015. How do labor market networks work? S. Kosslyn, S.
Richard, eds. Emerging Trends in the Social and Behavioral Sciences. John Wiley & Sons, 1-15.
Ryall, M.D., O. Sorenson. 2007. Brokers and competitive advantage. Manage. Sci. 53(4) 566-
583.
Sandberg S. 2013. Lean in: Women, work, and the will to lead. Alfred A. Knopf, New York.
37
Shih J. 2006. Circumventing discrimination: Gender and ethnic strategies in Silicon Valley.
Gend. Soc. 20(2):177–206.
Shipilov A, Labianca G, Kalnysh V, and Kalnysh Y. 2014 Network-building behavioral
tendencies, range, and promotion speed. Social Networks. 39: 71-83.
Silva JMCS, Tenreyro S. 2006. The Log of Gravity. Review of Economics and Statistics. 88(4)
641-658.
Sterling AD. 2014. Friendships and search behavior in labor markets. Manage. Sci. 60(9) 2341-
2354.
Sterling A.D. 2017. Taking the longer route: Tryouts and the search behavior of structurally
disadvantaged job-seekers.
Storvik, A., P. Schøne. 2009. In search of the glass ceiling: Gender and recruitment to
management in Norway's state bureaucracy. British Journal of Sociology 59(4) 729-755.Turco
CJ. 2010. Cultural foundations of tokenism: Evidence from the leveraged buyout industry. Am.
Sociol. Rev. 75(6) 894-913.
West C, Zimmerman DH. 1987. Doing gender. Gend. Soc. 1(2) 125-151.
38
Table 1. Descriptive statistics.
Mean SD (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
(1) Female 0.33 0.47 1
(2) emailclick 4.81 9.40 0.10 1
(3) emailclick_f 1.39 2.90 0.25 0.79 1
(4) emailclick_m 3.42 7.09 0.04 0.97 0.63 1
(5) emailclick_vp 1.08 2.58 0.00 0.81 0.53 0.83 1
(6) emailclick_jr 3.73 7.44 0.13 0.97 0.81 0.93 0.65 1
(7) GMAT (std) -0.01 1.00 -0.21 0.03 -0.06 0.06 0.03 0.03 1
(8) Sponsored 0.02 0.15 -0.05 -0.08 -0.06 -0.08 -0.07 -0.08 -0.02 1
(9) Native English Speaker 0.61 0.49 0.04 -0.09 0.00 -0.11 -0.01 -0.11 0.06 -0.03 1
(10) Extraversion 3.39 0.82 -0.04 0.07 0.13 0.05 0.08 0.06 0.00 0.06 0.16 1
(11) Search Breadth 0.29 0.46 0.01 0.17 0.16 0.16 0.09 0.19 0.00 -0.04 0.03 -0.02 1
(12) On Campus Resident 0.52 0.50 0.17 0.08 0.06 0.08 0.10 0.06 -0.02 0.00 -0.06 0.07 -0.03 1
(13) U.S. Citizen 0.64 0.48 0.05 -0.09 0.00 -0.11 -0.02 -0.10 -0.08 0.01 0.71 0.17 -0.01 -0.05 1
(14) Log Prior Experience 1.74 0.33 -0.08 -0.01 -0.01 -0.01 0.01 -0.02 -0.02 -0.03 0.01 -0.11 -0.01 -0.03 -0.02
39
Table 2. Descriptive analysis of networking behaviors with alums by male and female job-
seekers, showing mean values (standard deviations in parentheses) and the p-values of simple,
one-tailed t-tests of whether female job seekers contacted more alums than male job-seekers, on
average.
Male Job-Seekers Female Job-Seekers p-value of difference
emailclick 3.99 (7.73) 6.5 (12.00) 0.0310
emailclick_f 0.88 (1.69) 2.42 (4.28) 0.0007
emailclick_m 3.10 (6.46) 4.08 (8.24) 0.1427
emailclick_vp 1.03 (2.75) 1.19 (2.19) 0.2707
emailclick_jr 2.96 (5.55) 5.31 (10.14) 0.0182
40
Table 3. Regressions of networking activity with all alums (Model 1) and with various sub-
samples of alums (Models 2-5) for the full sample of student job seekers.
(1) (2) (3) (4) (5)
DV: emailclick emailclick_f emailclick_m emailclick_jr emailclick_vp
Female 0.359* 0.822*** 0.171 0.443* 0.0550
(0.174) (0.195) (0.187) (0.189) (0.258)
Asian 0.0674 0.222 -0.00809 0.226 -0.474
(0.295) (0.309) (0.339) (0.284) (0.364)
Other Ethnicity -0.575+ -0.331 -0.655+ -0.390 -1.229***
(0.325) (0.339) (0.360) (0.335) (0.339)
GMAT (std) 0.0132 -0.0731 0.0603 0.0113 0.0190
(0.0818) (0.0705) (0.106) (0.0743) (0.164)
Sponsored -2.025** -1.132* -2.822** -1.711* -22.68***
(0.667) (0.494) (0.985) (0.675) (0.483)
Native English
Speaker
-0.552+ -0.171 -0.706* -0.506+ -0.710
(0.297) (0.291) (0.352) (0.269) (0.443)
Extraversion 0.231 0.360* 0.185 0.225 0.250
(0.180) (0.168) (0.197) (0.179) (0.220)
Search Breadth 0.613* 0.549* 0.633* 0.666** 0.434
(0.257) (0.223) (0.291) (0.258) (0.304)
Log Work
Experience
-0.0162 0.0780 -0.0566 -0.0199 -0.0155
(0.239) (0.338) (0.233) (0.253) (0.296)
Constant 0.791 -1.598 0.873 0.393 -0.234
(0.872) (1.009) (0.919) (0.889) (0.947)
Observations 256 256 256 256 256
Standard errors in parentheses + p < .10, * p < .05, ** p < .01, *** p < .001
41
Figure 1. Histogram of the total number of alums contacted per student, separated out by the
gender of the student. The distribution may be skewed slightly to the right for female, compared
to male, students (p = 0.084).
0.1
.2.3
.4
De
nsity
0 20 40 60 80 100emailclick
Female Male
42
APPENDIX
Table A1. Regressions of network mobilization activity with all alums (Model 1) and with
various sub-samples of alums (Models 2-5) for the full sample of student job seekers (as in Table
3), controlling for each student’s ex ante career interests.
(1) (2) (3) (4) (5)
DV: emailclick emailclick_f emailclick_m emailclick_jr emailclick_vp
Female 0.222 0.629** 0.0465 0.257 0.0873
(0.207) (0.200) (0.228) (0.214) (0.263)
Asian -0.0379 0.238 -0.165 0.128 -0.600
(0.318) (0.292) (0.364) (0.309) (0.378)
Other -0.501+ -0.247 -0.583+ -0.318 -1.118**
(0.289) (0.276) (0.343) (0.278) (0.366)
GMAT (std) -0.0134 -0.106 0.0380 -0.0133 -0.00469
(0.106) (0.0776) (0.134) (0.0957) (0.184)
Sponsored -2.007** -1.123* -2.794** -1.706* -15.50***
(0.697) (0.495) (1.007) (0.685) (0.531)
Native English
Speaker
-0.394 -0.0412 -0.546 -0.348 -0.554
(0.351) (0.291) (0.418) (0.315) (0.494)
Extraversion 0.224 0.327** 0.180 0.222+ 0.258
(0.140) (0.124) (0.158) (0.129) (0.210)
Search Breadth 0.566* 0.445* 0.625* 0.640** 0.360
(0.232) (0.214) (0.270) (0.231) (0.291)
Log Work
Experience
-0.0608 0.0876 -0.115 -0.0700 0.00613
(0.270) (0.358) (0.270) (0.260) (0.361)
Interest in
Consulting
0.546* 0.0140 0.737** 0.639** 0.218
(0.221) (0.318) (0.238) (0.217) (0.290)
Interest in
Finance
-0.204 -0.396+ -0.143 -0.361+ 0.378
(0.181) (0.241) (0.201) (0.185) (0.263)
Interest in
General Mgmt
-0.351 0.0321 -0.493+ -0.315 -0.491
(0.288) (0.474) (0.297) (0.277) (0.365)
Interest in
Human Res.
0.883** 0.602* 0.989** 0.875** 0.876
(0.325) (0.295) (0.358) (0.284) (0.540)
Interest in Info
Tech
0.382 0.255 0.424 0.359 0.505
(0.247) (0.261) (0.269) (0.226) (0.410)
Interest in
Marketing
0.108 0.528* -0.0458 0.208 -0.275
(0.209) (0.246) (0.225) (0.204) (0.330)
43
Constant 0.620 -1.700* 0.680 0.164 -0.486
(0.717) (0.789) (0.812) (0.697) (0.953)
Observations 256 256 256 256 256
Standard errors in parentheses + p < .10, * p < .05, ** p < .01, *** p < .001
44
Table A2. Regressions of network mobilization activity with all alums (Model 1) and with
various sub-samples of alums (Models 2-5) for the full sample of student job seekers (as in Table
3), controlling for the industry of each student’s eventual internship.
(1) (2) (3) (4) (5)
DV: emailclick emailclick_f emailclick_m emailclick_jr emailclick_vp
Female 0.308 0.666* 0.156 0.406+ -0.139
(0.233) (0.261) (0.250) (0.233) (0.363)
Asian 0.195 0.345 0.122 0.324 -0.323
(0.280) (0.276) (0.331) (0.267) (0.376)
Other -0.452 -0.229 -0.528 -0.294 -1.134*
(0.355) (0.346) (0.409) (0.342) (0.484)
GMAT (std) -0.107 -0.225* -0.0460 -0.0962 -0.153
(0.110) (0.0922) (0.133) (0.102) (0.184)
Sponsored -1.976** -1.293* -2.652** -1.649* -14.43***
(0.706) (0.572) (1.004) (0.701) (0.737)
Native English
Speaker
-0.504 -0.109 -0.666 -0.514+ -0.488
(0.334) (0.285) (0.413) (0.296) (0.504)
Extraversion 0.348 0.462** 0.305 0.315 0.457+
(0.213) (0.168) (0.235) (0.209) (0.236)
Search Breadth 0.688* 0.562* 0.737* 0.750** 0.500
(0.286) (0.241) (0.317) (0.277) (0.364)
Log Work
Experience
-0.497 -0.239 -0.600 -0.485 -0.587
(0.396) (0.526) (0.396) (0.431) (0.450)
Consulting 0.466 0.156 0.635 0.443 0.676
(0.428) (0.411) (0.505) (0.449) (0.500)
Energy -0.343 -0.944 -0.0696 -0.468 0.318
(0.529) (0.581) (0.623) (0.549) (0.986)
Financial
Services
0.640 0.173 0.849 0.308 1.848**
(0.482) (0.482) (0.557) (0.481) (0.580)
Government 0.959 -0.0900 1.334+ 0.680 1.858*
(0.640) (0.773) (0.693) (0.664) (0.756)
Manufacturing 0.728* 0.476 0.857* 0.632+ 1.275*
(0.364) (0.429) (0.398) (0.355) (0.607)
Media,
Entertainment,
Sports
0.818 0.700 0.907 0.745 1.209
(0.615) (0.661) (0.781) (0.589) (0.853)
Pharma,
Healthcare,
Biotech
0.888 0.688 0.995 0.664 1.861**
(0.546) (0.565) (0.679) (0.543) (0.662)
Real Estate 1.903 0.110 2.428+ 1.687 2.885*
(1.234) (1.236) (1.283) (1.251) (1.236)
45
Retail 0.771 0.716 0.785 0.552 1.835**
(0.535) (0.495) (0.619) (0.594) (0.618)
Technology 0.588 0.476 0.666 0.566 0.649
(0.502) (0.494) (0.575) (0.515) (0.606)
Constant 0.594 -1.683 0.601 0.442 -1.396
(0.946) (1.027) (1.030) (0.944) (1.240)
Observations 214 214 214 214 214
Standard errors in parentheses + p < .10, * p < .05, ** p < .01, *** p < .001
46
The analysis of help received.
Recall that in the survey, each student was presented with a list of alums that included
some alums who were known to have been emailclicked by the student and some who were
known not to have been emailclicked by the student. We then asked whether the student had ever
interacted with each alum and, if so, what kind of help the alum provided. Students were asked
about the following types of help: “provided general career advice,” “provided general advice
about how to find an internship,” “informed me about a job opening”, “gave feedback on my
resume, cover letter, etc.,” “introduced me to his/her contacts, “put in a good word for me at
his/her company”, “acted as a formal referrer for me”. We coded these outcomes using three
dummy variables, representing different levels of help: Any Help was coded as 1 if the student
reported receiving any type of help and 0 if the student indicated that the alum did not provide
help of any type; Costly Help was coded as 1 if the student received an “introduction,” a “good
word,” or a “referral” and as 0 otherwise18; and Referral was coded as 1 if the student received a
specific job referral and 0 otherwise. Note that Costly Help and Referral both imply that the alum
was willing to put his or her own reputation on the line (Smith 2005) for the student; as such,
these are potentially costlier for the contact.
Then, using data on all the alums that the student contacted, we estimated dyad-level,
random-effects linear probability models of helping behavior with errors clustered on the
student. Key covariates include the student’s gender (Models 6-11), the alum’s gender (Models
9-11) and their interaction (Models 12-14). We employed random, rather than fixed, effects
because a student’s gender is constant across all alums with whom the student interacted. In
these models, we have 580 student-alum pairs that are clustered under 181 students. Note that
18 Our interview with the school’s director of career development indicated that an “introduction,” a “good word,” or
a “referral” were the types of help that are most significant and most valuable to students.
47
effects could not be identified for students who did not have variation on dependent variables
(i.e. received help from all contacts or did not receive help from any contacts), so these
observations were not included in this analysis.
Results in Table A3 indicate that the gender of the student has no statistically significant
effect on the likelihood of receiving any help, costly help or a referral. Though parenthetical to
the present study, we also note that for all three levels of help, we find no statistically significant
differences in the amount of help female alums gave to male and female students. Thus, these
results are inconsistent with a mechanism in which women contact more alums because they
receive less help from each alum contacted.
48
Table A3. Random effects linear probability models with errors clustered by student predicting the level of help a student received as
a function of the student’s and the alum’s gender.
(6) (7) (8) (9) (10) (11) (12) (13) (14)
DV: Any Help Costly Help Referral Any Help Costly Help Referral Any Help Costly Help Referral
Female Student -0.0184 0.0283 -0.00462 -0.0111 0.0242 -0.00398 -0.0244 0.0458 -0.0128
(0.0398) (0.0514) (0.0173) (0.0397) (0.0511) (0.0171) (0.0421) (0.0577) (0.0176)
Female Alum -0.0679* -0.00235 -0.00324 -0.0845+ 0.0243 -0.0138
(0.0342) (0.0376) (0.0158) (0.0438) (0.0488) (0.0196)
Female Student × Female Alum 0.0471 -0.0762 0.0306
(0.0679) (0.0764) (0.0338)
Asian 0.0207 0.0301 -0.0281 0.0249 0.0325 -0.0281 0.0251 0.0328 -0.0281
(0.0423) (0.0651) (0.0239) (0.0424) (0.0652) (0.0239) (0.0426) (0.0655) (0.0237)
Other 0.0208 0.0713 -0.0420 0.0221 0.0726 -0.0422+ 0.0247 0.0701 -0.0409
(0.0740) (0.0893) (0.0256) (0.0736) (0.0893) (0.0256) (0.0746) (0.0895) (0.0253)
GMAT (std) -0.00614 -0.0151 0.00447 -0.00755 -0.0157 0.00444 -0.00684 -0.0169 0.00485
(0.0139) (0.0229) (0.00619) (0.0140) (0.0229) (0.00624) (0.0140) (0.0232) (0.00642)
Sponsored 0.0623+ -0.0678 -0.0526** 0.0679+ -0.0669 -0.0524** 0.0669+ -0.0642 -0.0530**
(0.0355) (0.119) (0.0176) (0.0357) (0.119) (0.0173) (0.0356) (0.120) (0.0173)
Native English Speaker 0.0889+ -0.0103 -0.0400 0.0885+ -0.0107 -0.0401 0.0871+ -0.00772 -0.0413
(0.0494) (0.0592) (0.0266) (0.0497) (0.0592) (0.0266) (0.0497) (0.0601) (0.0267)
Extraversion 0.0143 0.0667** 0.0175* 0.0143 0.0668** 0.0175* 0.0136 0.0678** 0.0171*
(0.0168) (0.0250) (0.00841) (0.0168) (0.0250) (0.00841) (0.0167) (0.0252) (0.00823)
Search Breadth 0.00230 0.0121 -0.0150 0.00650 0.0138 -0.0149 0.00876 0.0101 -0.0134
(0.0282) (0.0491) (0.0139) (0.0284) (0.0489) (0.0140) (0.0281) (0.0495) (0.0142)
Log Work Experience -0.0586 -0.143+ 0.0126 -0.0546 -0.145+ 0.0129 -0.0527 -0.148+ 0.0140
(0.0555) (0.0785) (0.0502) (0.0542) (0.0786) (0.0502) (0.0543) (0.0795) (0.0502)
Constant 0.887*** 0.255 -0.00503 0.893*** 0.258 -0.00482 0.895*** 0.252 -0.00281
(0.136) (0.186) (0.0946) (0.135) (0.186) (0.0946) (0.134) (0.188) (0.0937)
Observations 580 580 580 579 579 579 579 579 579
Standard errors in parentheses + p < .10, * p < .05, ** p < .01, *** p < .001
Recommended